A Temporal Statistical Model for Entity-Relationship Schemas

This paper is about the statistical description of the contents of an existing or future database. This information, very often neglected in database design methods, is essential for database optimisation and physical tuning for instance. The paper defines an extension of the E-R model to specify the size of the populations of entity types, relationship types, domains, attributes, and of their interconnections. It proposes a minimum set of statistics together with the relations that organise them. It analyses the problem of computing these statistics by using the relations as derivation rules. Data statistics are then considered as time-dependent quantities. Finally, the paper proposes translation rules that allows the preservation of the statistics when their supporting schema is restructured. This statistical model has been implemented in a commercial CASE tool.